Book review: Algorithmic Regimes: Methods, Interaction and Politics (2024).
Juliane Jarke, Bianca Prietl, Simon Egbert, Yana Boeva, Hendrik Heuer, and Maike Arnold (Eds.). Amsterdam: Amsterdam University Press. 346 pp. ISBN 9789463728485 (E-Book)
Keywords Algorithms · Foucault · Algorithmic regime
Introduction
The purpose of the edited volume is to explore knowledge production under algorithmic regimes. Algorithmic regimes are, in this case, conceptualized as a combination of regimes of truth in a Foucauldian sense (Foucault 1977), and sociomaterial apparatuses in a Baradian sense (Barad 2007), something we find very generative and interesting. As such, we were initially thrilled to read and review this book. To us, it tackled an important gap in addressing the epistemological issues that emerge from ubiquitous use of algorithmic systems. We found the combination of a Foucauldian conceptual framework, including power/knowledge, governmentality, and regimes, together with Baradian concepts, such as intra-action, phenomenon and apparatus, deeply enticing and productive.
Foucault (1977, 1980) challenges traditional views of power, and proposes that it is instead produced through accepted forms of knowledge that shapes what is considered ‘true’, ‘right’, or ‘moral’ in a society (regimes of truth). These ‘truths’ can then be used to regulate and discipline people in a process of governmentality. The strategies and techniques used to govern people’s behaviour (or to make them gov- ern themselves) go beyond the traditional state and include institutions like schools, hospitals, and even ideas about how to live a good life. In a nutshell, power is not just about who has the upper hand; it is about the knowledge systems that shape our reality and how we are governed by them (power/knowledge). Today, many such knowledge systems are algorithmic in nature. Seeing how this is a key claim of the book, we find that Foucault’s ideas could have been introduced earlier on in this volume. As it is structured now, the chapters that properly define and make use of Foucault’s concepts are provided in the very last section of the book.
Furthermore, although Barad bases her theory on Foucault, she largely omits considerations of governance. Consequently, this book could exemplify a classic poststructuralist focus while still engaging with socio-material aspects. However, the initial promise to combine Foucault and Barad is not really fulfilled in the book. We think that Barad’s agential realism vocabulary has potential to create productive overlaps with Foucault’s concepts.
For example, Barad’s concept of intra-action describes the fundamental way in which actants in the world relate to each other, not as pre-existing entities that meet and bounce from each other, but through co-evolving, co-affecting, mutually transforming intra-actions that cause differences between phenomena to constantly get made, and unmade. A phenomenon is to Barad ‘… a specific intra-action of an “object”; and the “measuring agencies”; the object and the measuring agencies emerge from, rather than precede, the intra-action that produces them’ (Barad 2007: 128). These ‘measuring agencies’—or apparatuses—refer to the specific material- discursive practices that help to constitute phenomena by producing knowledge about them.
The key idea here is that apparatuses are not passive instruments of observation, they are material-discursive practices that create the differences and delimitations which are necessary for a phenomenon to be observed. So, apparatuses can be machines, but also other types of methods or conceptual frameworks, that are developed to ‘capture’ a certain phenomenon. This means that power/knowledge can be produced differently depending on the apparatus (e.g. algorithmic system) through which it is known. Unfortunately, none of the included chapters make use of these overlaps, and only a few make full use of Foucault’s framework.
Overall Structure of the Book
The book is divided into three sections: methods, interactions, and politics, each containing four chapters plus a ‘commentary’ providing summaries and perspectives on the included chapters. The introduction, ‘Knowing in algorithmic regimes’ (Jarke et al. 2024a, b), provides a brilliant launch of the book. It explores issues beyond the technical framework of algorithms, and brings in the cultural, political, social, eco- nomic, and other aspects that are always part of algorithmic governance, while also not forgetting their material entanglements.
The introduction is written in a way that provides an excellent overview of cur- rent issues and problems. It is lucid and introduces not only the chapters of the cur- rent volume, but also several other studies and books in a thoroughly accessible way, which serves as both an overview and an entry point into critical algorithm studies generally.
To us, the promised combined Foucauldian/Baradian approach is an evident and solid entry to critical studies of algorithms, and after reading the Introduction, we are even more convinced that it will be difficult for researchers to just disregard these theoretical perspectives in the future. The Introduction exemplifies the high- est standards of academic quality, and we earnestly hope it attracts the attention and recognition it deserves from both the academic community and a wider audience.
Methods
After the great introduction, the methods section begins with the chapter ‘Revisiting Transparency Efforts in Algorithmic Regimes’ (Eslami and Heuer 2024). The authors problematize transparency, user levels (primary, secondary and tertiary users), and public algorithms (algorithms that effectively belong to citizens) in interesting and thought-provoking ways. The concepts introduced here clearly have analytical traction. In many ways, the authors’ discussions of previous research are some of the main contributions of this chapter. However, the proposed solution that we need more knowledge and education could arguably be problematized in itself.
Increased individual knowledge of how algorithms work ‘under the hood’ is positioned as a neutralized and benevolent ‘solution’ for making these regimes more equitable or just. We would argue that, as users, we are already submerged in (capitalist or even feudalist) algorithmic regimes, and just knowing how they work might not be enough to change neither systemic injustices nor our behaviours or decisions. Many times, we are only given the option to ‘like, or do, other things’ in order to tweak the output from the algorithmic regime.
The following chapter (Gramelsberger et al. 2024) continues on the same track. The authors aim to establish a methodology called Computational Science Code Studies (CSS) and argue that traditional ethnographic studies are insufficient for thoroughly understanding computational practices. The central question they pose is: How can we make computational science more accessible? To answer this, they present software tools designed to analyse comments within computer code. In programming, comments are non-executable annotations that explain the code’s function to the reader. By examining these comments, the authors suggest that one can trace the genealogy of the code, revealing its evolution over time.
Additionally, their tools aim to visualize various functions within the program by highlighting the use of pre-made components in the code. However, we might question the novelty of this approach, as it appears similar to practices already employed by system engineers and programmers, particularly in debugging processes, that involve checking dependencies on specific libraries. This raises the question: How does this tool improve upon existing methods of interviewing programmers or con- ducting ethnographic studies? The key insight of the chapter is that for anything to be converted into computer code, it must be quantifiable. In other words, for some- thing to be processed by a computer, it must be expressed in computational terms. Who knew?
Chapter 4 by Storms and Alvarado (2024) provides a methodological contribution relating to how poor knowledge, or skewed preconceptions, can generate difficulties when researching people’s experiences or perceptions of what algorithms do/are. As readers, we are not entirely convinced by the idea that users have ‘erroneous algorithmic beliefs’ and that these then need to be ‘corrected’. The fact that ‘folk models’ and imaginaries are potentially part of any complex phenomena in people’s worldview is, to us, one of the most interesting preconditions of phenomenological studies. Also, regardless of whether people have a ‘more correct’ view of algorithms or not, the knowledge/power these regimes hold and execute will likely persist.
As such, Chapter 4 could have benefitted from engaging with the arguments presented in Chapter 5, where Jarke and Heuer (2024) contend that merely unpacking ‘black-boxed’ technology is limited and potentially misleading. Instead, the authors emphasize that viewing the black box of algorithmic machines as a mode of inquiry is more fruitful. The authors address the concept of black boxes in two ways: first, as devices that produce and record data, analogous to the black boxes in airplanes, which reflect back on the researcher and facilitate independent studies; and second, as systems that are inherently unknown or unknowable. The authors then employ the metaphor of folding to present a nuanced understanding of previous research, laying a solid foundation for future investigations.
The chapter excels in distinguishing what is novel from what has already been accomplished. The writing is clear and thorough, and the authors critically and reflectively engage with epistemic issues, demonstrating that merely revealing the technical details of the black box is insufficient. Such an approach would only limit the understanding of how algorithms affect individuals across different contexts, times, and spaces.
Jarke and Heuer (2024) further address the idea of the ‘monster’, but here we wonder if the notion of ‘capta’ (as in interpreted and constructed data) (Drucker 2011) or ‘situated data’ (Lavin 2021) might not be better terms? Everything out there is just messy data. So, maybe everything that is, in some sense, excluded, does not automatically qualify as a monster (it is still messy data, and potential capta)? Also, in a Foucauldian sense, monsters are understood as something absurd or abject—non-understandable—that disrupts the knowledge-power distribution, something that is impossible to make sense of (at least now). It is difficult to imagine what this is, seeing it is fundamentally unthinkable. Just that something is outside or irrelevant to a certain algorithm is not enough to make it a monster. However, this is the only small piece of criticism we identify in this little gold nugget of a chapter that we hope will be widely read.
The ‘Methods’ section concludes with a commentary by Mackenzie (2024). The inclusion of commentaries throughout the book is an excellent feature, providing insightful conclusions to the sections. These commentaries often summarize the content and sometimes offer additional contributions. Mackenzie’s (2024) commentary is particularly noteworthy for its constructive criticism of the preceding chapters, echoing some of the points we have raised, albeit in a more nuanced and considerate manner. His remarks significantly enhance the reader’s understanding by engaging in a dialogue with the chapters, examining concepts, and clarifying terms that were insufficiently explained, such as ‘regime’. This thoughtful approach is greatly appreciated and contributes to the overall depth and clarity of the book.
Interaction
The next section focuses on ‘interaction’. In its first chapter, authors Boeva and Kropp (2024) examine computational architectural design. Here we would have liked to see a clearer historical acknowledgment of authors dealing with similar issues (e.g., Wajcman 2006; Margolin 1995; Cockburn 1981)—what can we take away from these, much older, studies when studying algorithmic regimes in relation to architectural design?
In Chapter 8, Büchner et al. (2024) argue that organizations must be understood as social systems that enable and restrict how algorithmic systems unfold. Their analysis focuses on three dimensions: goals, differentiations, and goal conflicts. In conclusion, the authors find that ‘organizations enable and simultaneously restrict, break and relativize the power of algorithmic regimes’ (Büchner et al. 2024: 178). While this may seem obvious, we also find that fundamental insights like these con- tribute to making this book a good introductory textbook for students.
This feeling is strengthened by the next chapter, on trust regimes and fake news, by Wiengarn and Arnold (2024). This chapter provides a good, albeit sometimes conflicting (in terms of whether trust is a systemic or individual issue), overview of trust in relation to news (which would, again, serve nicely as an introduction for students). A question that comes to mind, however, is why being deceived by technology is necessarily a trust (and value) issue, and not more a medium-specific or media-logical issue? Issues around social hierarchies and profit in journalism could have been pertinently included as well.
Next is a chapter that stands out as a highlight in the book. Poechhacker et al. (2024) adeptly draw on Dewey’s concept of ‘issue publics’ to argue that various algorithmic techniques for recommendations significantly influence the formation of publics by mediating practices within an algorithmic regime. Typically, algorithmic filter systems are criticized for segregating ‘publics’ into echo chambers and filter bubbles. However, the authors challenge this perspective by demonstrating that these criticisms rest on two assumptions: that filter bubbles inherently exist, and that a uniform public sphere is an essential precondition. They argue that these assumptions may not hold true, and perhaps never have. Classical political theory lacks the tools to address how publics evolve through algorithmic techniques, highlighting the need for new theoretical frameworks to understand information distribution, its scope, and its limits.
The authors provide a thorough genealogy of filters, information overload, and the early aspirations (and possible nightmares) of the Internet. This historical context deepens our understanding of the complexity behind current algorithmic solutions and their emergence. Moreover, the authors make a significant contribution to knowledge by illustrating how ideal types of recommendation algorithms can be integrated with pragmatic theories of democracy and publics. They explore whose actions are permitted to participate in these spaces and how those actions are subsequently processed by algorithms. This chapter not only advances theoretical discussions but also offers practical insights into the dynamic inter- play between algorithmic techniques and publics.
The commentary on the ‘interaction’ section is provided by Milan (2024). Here, we are provided with interesting ideas on how algorithms define ‘realities’, and that they often follow a logic of ‘trial-and-error’, where algorithms are continuously tested on groups (often already marginalized or exposed groups), and gradually ‘improved’, creating yet another version of ‘reality’. As a commentary, it is a bit different from the previous one (Mackenzie 2024) as Milan’s (2024) piece is more of a reflection rather than a synthesizing summary. Nevertheless, the idea of having commentaries of this kind is very nice, and the fact that they differ in style may be considered a strength.
Politics
The final section of the book concerns politics. It is in this section that Foucault’s concepts of ‘regimes’ and ‘governmentality’ are more thoroughly explained and put to use (previously they have mostly been hinted at). As such, it would perhaps have been a good idea to put this section first in the book.
In the first chapter in this section, Prietl and Raible (2024) present a study on the institutionalization of academic data science in the DACH region. The authors look at data science as constitutive for new algorithmic regimes of knowledge production. They write: ‘In this sense, data science represents a specific set of knowledge production techniques and, thus, a form of power, as well as an instrument that can promote different interests and support different power relations’ (Prietl and Raible 2024: 243). They show how data science aligns with, for example, economics or life sciences, in order to become applicable. As such, data science is presented as a universal toolbox, harvesting data from the ‘real world’, which also needs to be coupled with an ‘application domain’ in order to provide, what is often presented as, neutral, objective and actionable science. This is an excellent chapter which highlights how data science is a field that claims to provide universal solutions, while also being ‘just a tool’. As such, the epistemological foundations of data science as a field are highlighted in a thoroughly interesting way.
The following chapter, ‘Algorithmic Futures: Governmentality and Prediction Regimes’ (Egbert 2024), is also an important contribution of the book. The chapter offers a brilliant discussion of Foucault and governmentality in relation to predict- ability. Simon Egbert argues, with the help of Rouvroy, that technologies function as ‘rendering devices’ that facilitate governmental practices. We see it as relating to Haraway’s (1998) assertion that everything that can be codified will be codified, and that Egbert (2024) extends this idea to a higher level of abstraction; everything that can be controlled must be controlled. This chapter provides profound insights into the intersection of technology, governance, and control, making it an essential read for understanding the implications of algorithmic prediction regimes.
In ‘Power and Resistance in the Twitter Bias Discourse’, Lopez (2024) reports on a case of racial bias in Twitter’s image cropping tool, which in turn sparked a conversation about algorithmic discrimination on the platform. Twitter’s response, to hold a contest where users were encouraged to come up with ideas on how to improve the algorithm, can be seen as a way to manage the discourse and potentially control how users identify and discuss bias. The chapter explores this as a complex interplay of power, but does not really manage to extend the discussions further.
In ‘Making Algorithms Fair: Ethnographic Insight from Machine Learning Interventions’, Kinder-Kurlanda and Fahimi (2024) examine algorithmic regimes from the perspective of those who want to build better algorithms. Personally, we look forward to the next text, based on the interesting questions that the authors ask towards the very end of the text.
The final commentary by Thylstrup (2024) raises, for us, questions around the necessity to invent new concepts to address algorithmic regimes. Thylstrup’s brief introduction of the concept of the ‘gimmick’ is inspiring however, and we would have loved to see that further developed.
Old Frameworks, New Perspectives
A strength demonstrated by the book is that the re-examination of existing conceptual frameworks can provide fresh perspectives on complex contemporary issues like algorithms and the rise of Artificial Intelligence. As shown from our walkthrough, certain chapters accomplish this better than others. But the remaining question is: Do we need another book on algorithms and their regimes? Hasn’t enough been written about it? Do the chapters live up to the introduction? And who are the ‘we’ that may need this book? We find that the book can fill a place as a handbook for students and we can see its use in different types of courses. As a whole, however, and for every chapter to contribute new perspectives, the book is a bit too uneven. Having said that, the outstanding chapters in this book offer insightful perspectives beyond the usual algorithm hype.
So, while the book highlights the continued relevance of older theories, it does not always delve deeply enough into how they might need to be adapted or expanded to address truly novel aspects of contemporary technology. Several chapters lack innovation (other than in an empirical sense) and the analyses come across as some- what shallow or repetitive.
In many ways, this book offers a survey of how established theories from Science and Technology Studies (STS), Human–Computer Interaction (HCI), or Poststructuralism can be applied to analyse contemporary technological phenomena. While the demonstration that these older theories remain relevant certainly holds merit, this is also where the book’s execution falls short. The book would benefit from acknowledging and potentially integrating newer theoretical frameworks alongside the more established ones.
Algorithmic Regimes: Methods, Interaction and Politics (Jarke et al. 2024a, b) offers a great starting point for understanding how existing theories can be applied to contemporary technology. However, its focus on established ideas and textbook- like presentation restricts its potential impact. Despite those limitations, there are also real gems in the book, several chapters that exhibit exceptional erudition and offer novel perspectives that transcend the constraints of its overall structure. These sections, characterized by their lucid and insightful prose, counterbalance the less compelling portions of the text, thereby elevating the work’s overall scholarly merit.
Book Review: Digital Roots – Historicizing Media and Communication Concepts of the Digital Age
Gabriel Balbi, Nelson Ribeiro, Valérie Schafer and Christian Schwarzenegger (eds). (Berlin/ Boston: De Gruyter, 2021), pp. 318. ISBN: 978-3-11-073988-6.
This review was published in Rundfunk & Geschichte, 2022(1-2): 115-117.
This ambitious volume goes through sixteen concepts in order to outline a “conceptual media and communication history”. In doing this, the book also seeks to avoid ‘presentism’ (i.e. it confronts the alleged newness of certain concepts), and scrutinize how concepts both persist and develop over time. The concepts, with each one covered in one dedicated chapter, are: networks; convergence; multimedia; interactivity; artificial intelligence; global governance; datafication; fake news; echo chambers; activism; amateurism; digital loneliness; telepresence; user-generated content; fandom; and authenticity.
The main argument of the book is that digital media concepts have roots and far reaching complex historical connections, which are today often overlooked. The editors of the books also aim to set a new trend in research by opposing the often told teleological stories of digital media innovation. Thereby, the book shows kinship to media archeological ambitions. Nevertheless, many of the included chapters also retain a certain linear historication and chronology of ‘classic works’ (both in terms of previous theoretical work and empirical examples). This goes back to the overall aim of the book, which includes to “identify the origins, sources, lineage and heritage of some of the most mental images of the digital world” (p. 2). The ambition of several chapters thus becomes more a question of showing that concepts in fact have a pre-digital history, rather than to identify potentially overseen empirical phenomena, or to effectively illustrate the concepts’ usefulness today.
Indeed, the recurring emphasis on the argument that the pre-digital roots of these concepts are overlooked arguably comes across a straw-man argument (i.e. the idea that there is a need to convey that concepts we use today are not unprecedented and entirely novel), at least to scholars who have been researching computer-mediated communication since the early days of the Internet (and even before). To them, many chapters will recount familiar (while materially focused) historical descriptions, rather than analyses or developments of new analytical nuances or entry points. At times, this is due to the fact that the book has such a clear focus on media studies. The relation between disciplines such as HCI and media and communication science comes across as one where media and comms are only now discovering many of the issues that have been pondered by (the more philosophical parts of) computer science (including the many humanities and social science scholars who have been part of it) for decades (and presenting them as new, albeit for a new audience). The ambition is thus set high, but several chapters in fact fail to meet these expectations.
This means that many of the concepts have much more widespread theoretical and multidisciplinary roots than they are given credit for. That is, several of the concepts covered in this volume have been integral to other sciences for a long time (discussions which are at times ignored). So, in order to not “reinvent the wheel”, such previous efforts could also have been acknowledged and incorporated. As mentioned, the book as a whole may have a too stringent focus on (certain) media classics, which is great for media students (although some would argue that too many concepts are missing for this book to be really useful as an introductory to media studies), but for others it will come across as unnecessarily delimited. Works from Science and Technology Studies could easily have been included, and potentially also helped to argue against a teleological media perspective. One could thereby argue that the book displays a kind of active amnesia which could be interpreted as a way to retain disciplinary boundaries. As such, the book may provide potential readers with very little new knowledge about digital roots, and run the risk of cementing a history of important events and persons, rather than challenge teleology, or make clear parallels to our contemporary media ecology. The result is, contrary to its ambitious purpose, often a chronological description that repeats a, to many, well-known history. Arguably, there is both a familiarity to, as well as an arbitrariness in, the examples – caused by the fact that few authors explain why they highlight the examples they do (and not others). This, again, makes us wonder who the imaginary reader is; who needs to be convinced that these media concepts are not new, or that they lack complex historical roots?
Having said that, there are certainly chapters in the book that are particularly strong. These six chapters go beyond just showing that concepts have a pre-digital history, and dives deeper into specific historical data and/or contribute with a new theoretical facet of a specific concept.
One such chapter is Day Good’s chapter on multimedia, which provides one of the more empirically driven, and most interesting, chapters of the book. Good’s concept of ‘media litanies’ gives the reader not only an interesting “sidetrack” (much in the line of media archeological research), but also a rich insight into early multimedial combinations and imaginaries (e.g. “illustrated radio”). It does not stop there however, but also considers the broader societal consequences of these technologies, as well as discusses how they intermingle with political ideologies. The chapter contains a stringent analysis of how visions about media and pedagogy overlap and influence what media technologies that are actually deployed in schools. Day Good effectively shows how schooling and ed-tech are always already entangled, and what such entanglements have meant at different points in time.
Another chapter, warranting the purchase of the book, is Digital Media Activism (by Treré and Kaun). Here, the authors eloquently show that the border between digital and non-digital is always permeable (i.e. all activism is hybrid). This chapter provides a knowledgeable and thorough overview of the field. Treré and Kaun make sure to theorize and problematize, and not just describe, effectively demonstrating how a reconsideration of history can provide new perspectives.
The chapter on datafication further contributes with useful analytical terms. This chapter fruitfully considers ‘what is at stake’ when we talk about datafication. Again, it goes beyond being descriptive and just showing that the concept has a history. Koenen, Schwarzenegger and Kittler politicize the concept, and thereby demonstrate an interesting analysis of “[p]ersistent questions and changing answers through the ages” (p.152). We would happily read a longer exposition of how, not only ‘dark data’ emerge, but also how it could be complemented by ‘light data’. That is, data has undoubtedly been important for the construction of the welfare state (i.e. a redistribution of resources requires some data as a ground for decisions and policies). For example, one could also have noted how the idea of ‘nurturing’ the population (and not just exploit it) through the use of data (e.g. health data, nativity data, mortality data, social stratification data, and education data) gave birth to progressive changes in legislation, and became an incredibly important tool to govern populations in other directions as well. A nuanced discussion on how light and dark data can converge into new forms of biopolitical governance would have strengthened the contribution even further.
Further, Bourdon’s chapter on telepresence is a pleasant read and presents original insights into how the concept addresses affective and emotional relations through, and with, media. For readers of this journal, this chapter has the most clear emphasis on broadcasting. There is also a certain overlap with the following chapter on digital loneliness, where author Brennan also provides thoughtful ideas on how we live in an individualized and capitalist system that practically promotes loneliness: “Loneliness might then be driven, not by technology, but by the commercial colonization of everyday life.” (p.240). As such, people are pushed into a ubiquitous competitive mode, where just enough individuality (but not too much) is rewarded. Constant demands for authenticity and conformity brush up against constant comparison and adaptability. As Brennan puts it: “There is an absence of any kind of personal quality that cannot be changed in response to the market” (p. 241). In this paradoxical and anxious reality, Brennan interestingly proposes that scripted technological relations might in fact be perceived as less risky compared to commercially driven and competitive human relations.
Finally, the chapter on fandom by Benecchi and Wang is also one of the book’s best chapters. It provides concrete, and seldom discussed, empirical examples, providing useful insights into broadened contexts (social, cultural, economic). The reader is now (at the end of the book) also faced with how many of the previous chapters (and much of research in general) has been focused on a western context. This is a vital contribution as it makes the reader rethink and hopefully revisit the other chapters once more. In this way, and in line with the aims of the authors, the chapter helps to destabilize the anglophone focus that media studies has had, and continues to have.
In summary, the book takes on an important task in untangling the roots of central concepts used in media science. We do agree that there is a continuous need to be reminded of the non-newness and roots of theoretical concepts, and in this way the book clearly shows that these concepts are not novel, and have a long(er) history. At the same time, we would have liked to have seen an elaboration and motivation on which concepts that were included (and why) and which were discarded. As it stands now, the book has a certain skew towards concepts that are currently outdated (or at least out of fashion). One way to follow the book’s ambitions more thoroughly could have been to deliberately address more modern concepts, and make even more use of the term roots, and focus on, for example, rhizomes or from where these concepts receive their ongoing nutrition (from various sources)? How have deeper and broader structures provided preconditions that allow for certain concepts (or media technologies) to grow, and others to whither? Why do certain technologies emerge when and where they do? Or even asking how a digital and conceptual photo-synthesis works? In this respect, we find that a number of chapters (the ones mentioned above) stand out and warrant a purchase of the book. These six chapters in particular offer new and deeper insights into conceptual roots both for media scholars as well as researchers in other fields.